docs: replace agent-rdp skill with validated workflow - #79
Open
luckygreen wants to merge 1 commit into
Open
Conversation
Replace the shipped skill with an OS-neutral, path-neutral, and harness-neutral workflow containing only live-tested agent-rdp 0.6.5 surfaces. Correct screenshot, scrolling, session, and keyboard guidance while recording the exact Windows 11 and Ubuntu WSL validation environments. Validated with the official skill validator and live RDP tests against Windows 11 Pro 24H2 build 26100.8655.
denisix
pushed a commit
to denisix/agent-rdp
that referenced
this pull request
Aug 9, 2026
Merged from thisnick#79 (luckygreen), with two fixes applied on top: - Restored the allowed-tools: Bash(agent-rdp:*) frontmatter scoping that the original PR dropped. - Restored the Claude-specific warning against estimating click coordinates from screenshots, which the rewrite otherwise removed. Replaces the shipped skill with a path-neutral, harness-neutral workflow derived from live testing of agent-rdp 0.6.5, distinguishing confirmed CLI/Node API/WebSocket behavior from documented-but-broken commands (obsolete screenshot --base64, keyboard key vs press, positional scroll arguments, session close, clipboard init blocking).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Replace the shipped agent-rdp skill with a path-neutral, harness-neutral, and operating-system-neutral workflow derived from live testing of agent-rdp 0.6.5.
The replacement distinguishes confirmed CLI, Node API, WebSocket, web-viewer, RDP, clipboard, drive-mapping, screenshot, OCR, mouse, keyboard, and UI Automation behavior from commands that are documented but do not work as written.
Scope
This pull request changes only
skills/agent-rdp/SKILL.md.It does not modify runtime code, package metadata, examples, release automation, or unrelated documentation.
Tested Environment
The native Windows controller was Windows 11 Pro 24H2, 64-bit, OS build
26100.8875, running PowerShell 7.The Linux controller was Ubuntu 24.04.4 LTS, Noble Numbat, running under WSL 2.7.11.0 with kernel
6.18.33.2-microsoft-standard-WSL2.The target was Windows 11 Pro 24H2, 64-bit, OS build
26100.4652.The npm package under test was agent-rdp
0.6.5. The inspected upstream source was commite4c45f9c4ec2a01694c148fd51d358d7277ad42a.Confirmed Passing Behavior
Ordinary RDP connection, PNG screenshots, mouse point and click, right click, double click, move, drag, Unicode keyboard input, key combinations, positional scrolling, OCR, clipboard transfer after initialization, drive mapping, named sessions, JSON output, the web viewer, the WebSocket protocol, and the Node API passed.
UI Automation also passed after allowing the newly connected Windows desktop to stabilize before bootstrap.
Corrections Captured by the Skill
The CLI writes screenshots to disk and does not support the obsolete
screenshot --base64flag.PNG screenshots pass. CLI JPEG capture fails when the decoded frame is RGBA because the JPEG encoder rejects the alpha channel.
Scrolling uses positional arguments. The documented
--amountform is invalid.Keyboard key operations use
keyboard press. The documentedkeyboard keyform is invalid.The CLI has no
session closesubcommand.Clipboard operations can block during first-use initialization until the remote clipboard channel completes setup.
The Node API can leave the process alive after disconnect unless the caller performs explicit cleanup.
Immediate UI Automation bootstrap can fail when the remote desktop is not ready to accept input. A stabilization delay produced a passing bootstrap on the tested target.
Relationship to Existing Pull Requests
Open PR #74 overlaps the UI Automation bootstrap timing problem by switching bootstrap command entry to clipboard paste and increasing the delay after
Win+R. It is a partial match, not a complete resolution of the tested five-second stabilization requirement.Open PR #78 removes obsolete
screenshot --base64documentation from the README and example script. It does not update the shipped skill or address the remaining tested discrepancies.The legacy Windows RDP TLS change in the
kfuglsangfork is intentionally excluded. The tested target connected successfully, and every relevant failure occurred after connection.Validation
The proposed skill passes the repository's skill validator.
The generated change applies cleanly to upstream commit
e4c45f9c4ec2a01694c148fd51d358d7277ad42a.The skill contains no controller-specific absolute paths, no harness-specific installation paths, no bullet lists, and no en dash or em dash characters.
Authorship
Author: luckygreen (with assistance by 5.6 Sol High)